Research on video image stabilization decision algorithm based on ROI block matching method

Abstract

Aiming at the problem that the feasibility and accuracy of image measurement and analysis are reduced due to video stabilization in automobile collision experiment, a video image stabilization decision algorithm based on block matching method based on region of interest (ROI) is proposed in this paper. Firstly, in the first frame, the user selects the ROI region containing the marker identification, uses the block matching method to automatically match the ROI region features frame by frame, and estimates the motion of the video frame. And then determines whether the feature matching fails according to the motion of the video frame. When the feature matching fails, the search range can be reduced and the ROI features can be re matched to improve the matching accuracy and speed, Finally, the video frame sequence after image stabilization is output by frame reverse motion compensation according to the amount of motion. The experimental results show that the proposed algorithm can effectively reduce the video stabilization in collision experiments in the subjective evaluation and the objective evaluation of local peak signal-to-noise ratio.